PPubMed30 Nov 2016Cited 2.1K×

Sensitivity Analyses for Robust Causal Inference from Mendelian Randomization Analyses with Multiple Genetic Variants

Stephen Burgess, Jack Bowden, Tove Fall, Erik Ingelsson, Simon G. Thompson

Abstract

'Simon G. Thompson'] Supplemental Digital Content is available in the text.

A figure from Sensitivity Analyses for Robust Causal Inference from Mendelian Randomization Analyses with Multiple Genetic Variants — fig. from the paper

§ The Valyu brief

The paper, in five minutes

Read from the full text via Valyu Contents

The paper outlines sensitivity analyses and robust methods to assess and improve causal inference in Mendelian randomization using multiple genetic variants.

What’s new

i.Describes a suite of sensitivity analyses for MR with multiple variants using summarized data
ii.Introduces robust analysis methods (penalization, median‑based, Egger regression) that relax standard IV assumptions
iii.Applies these methods to CRP–CAD example, illustrating heterogeneity, pleiotropy, and leave‑one‑out analyses
iv.Provides guidance on using measured covariates and gene‑environment interactions to detect invalid instruments
v.Discusses limitations of MR assumptions and the interpretability of causal estimates

How they did it

The authors reviewed and compiled existing sensitivity‑analysis techniques, then applied them to publicly available summary statistics for C‑reactive protein (CRP) genetic variants and coronary artery disease risk. They performed heterogeneity testing (Cochran’s Q), funnel‑plot inspection, Egger regression, and systematic leave‑one‑out analyses using inverse‑variance weighted estimates, all based on summarized data.

What they found

Cochran’s Q statistic = 71.9 with 16 degrees of freedom, indicating substantial heterogeneity
P‑value for directional pleiotropy from Egger regression = 0.61 (no evidence of pleiotropy)
Inverse‑variance weighted estimate using 17 genome‑wide variants was negative overall but became positive after omitting two strongest negative‑effect variants
Scatter plots showed homogeneity for CRP‑region variants and heterogeneity for genome‑wide variants

Caveats

✕Instrumental variable assumptions (especially no association with unmeasured confounders) cannot be fully tested
✕Gene‑environment interaction analyses may suffer from collider bias when stratifying on the risk factor
✕Robust methods such as Egger regression may lose consistency properties under certain violations
✕Sensitivity analyses cannot pinpoint the true causal risk factor when a surrogate (e.g., BMI) is used

§ The Valyu brief

Reading the full paper and taking notes. This takes a few seconds…

§ Ask this paper

Ask a question about this paper

Valyu reads the full text and answers from what the paper actually says.

§ Related work in other archives

Sensitivity Analyses for Robust Causal Inference from Mendelian Randomization Analyses with Multiple Genetic Variants

The paper, in five minutes

Ask a question about this paper

Similar papers in all four archives

Similar papers in all four archives

The paper, in five minutes